Multiscale uncertainty quantification for foam-filled GFRP columns under axial compressions

Abstract

This study proposes a novel semi-analytical multiscale framework to quantify the ultimate axial load of foam-filled GFRP columns under mixed uncertainties across micro- and macro-scales. A key challenge is to capture random and interval uncertainties while maintaining computational efficiency. Traditional deterministic models ignore uncertainty, whereas high-fidelity Monte Carlo simulations are computationally prohibitive. To address this gap, the framework integrates the Mori–Tanaka homogenization method with a hybrid uncertainty propagation strategy that combines perturbation-based probabilistic analysis for random variables and particle swarm optimization for interval variables. Validation against Monte Carlo benchmarks confirms high accuracy. The relative errors in the predicted mean and standard deviation of the ultimate load are within 2%. Importantly, the method achieves this precision while reducing the computational cost by several orders of magnitude. This development turns uncertainty quantification from a computationally intensive process into a practical engineering tool. It enables efficient parametric studies and reliability-based design optimization of composite structures.